Skip to content

Unable to make imatrix (and likely quant) for nvidia's ChatQA-1.5 8B #7046

@bartowski1182

Description

@bartowski1182

This model: https://huggingface.co/nvidia/ChatQA-1.5-8B

Conversion worked no issue, but then when it's time to calculate the imatrix I see:

llama_model_load: error loading model: done_getting_tensors: wrong number of tensors; expected 323, got 291

Doing a search, last time slaren mentioned doing a gguf-dump.py, here's the output:

* Loading: /models/ChatQA-1.5-8B-GGUF/ChatQA-1.5-8B-fp16.gguf
* File is LITTLE endian, script is running on a LITTLE endian host.
llama_cpp-1  |
* Dumping 24 key/value pair(s)
      1: UINT32     |        1 | GGUF.version = 3
      2: UINT64     |        1 | GGUF.tensor_count = 323
      3: UINT64     |        1 | GGUF.kv_count = 21
      4: STRING     |        1 | general.architecture = 'llama'
      5: STRING     |        1 | general.name = 'ChatQA-1.5-8B'
      6: UINT32     |        1 | llama.block_count = 32
      7: UINT32     |        1 | llama.context_length = 8192
      8: UINT32     |        1 | llama.embedding_length = 4096
      9: UINT32     |        1 | llama.feed_forward_length = 14336
     10: UINT32     |        1 | llama.attention.head_count = 32
     11: UINT32     |        1 | llama.attention.head_count_kv = 8
     12: FLOAT32    |        1 | llama.rope.freq_base = 500000.0
     13: FLOAT32    |        1 | llama.attention.layer_norm_rms_epsilon = 9.999999747378752e-06
     14: UINT32     |        1 | general.file_type = 1
     15: UINT32     |        1 | llama.vocab_size = 128256
     16: UINT32     |        1 | llama.rope.dimension_count = 128
     17: STRING     |        1 | tokenizer.ggml.model = 'gpt2'
     18: STRING     |        1 | tokenizer.ggml.pre = 'llama-bpe'
     19: [STRING]   |   128256 | tokenizer.ggml.tokens
     20: [INT32]    |   128256 | tokenizer.ggml.token_type
     21: [STRING]   |   280147 | tokenizer.ggml.merges
     22: UINT32     |        1 | tokenizer.ggml.bos_token_id = 128000
     23: UINT32     |        1 | tokenizer.ggml.eos_token_id = 128001
     24: STRING     |        1 | tokenizer.chat_template = '{% set loop_messages = messages %}{% for message in loop_mes'
llama_cpp-1  |
* Dumping 323 tensor(s)
      1:  525336576 |  4096, 128256,     1,     1 | F16     | token_embd.weight
      2:       4096 |  4096,     1,     1,     1 | F32     | blk.0.attn_norm.weight
      3:   58720256 | 14336,  4096,     1,     1 | F16     | blk.0.ffn_down.weight
      4:   58720256 |  4096, 14336,     1,     1 | F16     | blk.0.ffn_gate.weight
      5:   58720256 |  4096, 14336,     1,     1 | F16     | blk.0.ffn_up.weight
      6:       4096 |  4096,     1,     1,     1 | F32     | blk.0.ffn_norm.weight
      7:    4194304 |  4096,  1024,     1,     1 | F16     | blk.0.attn_k.weight
      8:   16777216 |  4096,  4096,     1,     1 | F16     | blk.0.attn_output.weight
      9:   16777216 |  4096,  4096,     1,     1 | F16     | blk.0.attn_q.weight
     10:         64 |    64,     1,     1,     1 | F32     | blk.0.attn_rot_embd
     11:    4194304 |  4096,  1024,     1,     1 | F16     | blk.0.attn_v.weight
     12:       4096 |  4096,     1,     1,     1 | F32     | blk.1.attn_norm.weight
     13:   58720256 | 14336,  4096,     1,     1 | F16     | blk.1.ffn_down.weight
     14:   58720256 |  4096, 14336,     1,     1 | F16     | blk.1.ffn_gate.weight
     15:   58720256 |  4096, 14336,     1,     1 | F16     | blk.1.ffn_up.weight
     16:       4096 |  4096,     1,     1,     1 | F32     | blk.1.ffn_norm.weight
     17:    4194304 |  4096,  1024,     1,     1 | F16     | blk.1.attn_k.weight
     18:   16777216 |  4096,  4096,     1,     1 | F16     | blk.1.attn_output.weight
     19:   16777216 |  4096,  4096,     1,     1 | F16     | blk.1.attn_q.weight
     20:         64 |    64,     1,     1,     1 | F32     | blk.1.attn_rot_embd
     21:    4194304 |  4096,  1024,     1,     1 | F16     | blk.1.attn_v.weight
     22:       4096 |  4096,     1,     1,     1 | F32     | blk.10.attn_norm.weight
     23:   58720256 | 14336,  4096,     1,     1 | F16     | blk.10.ffn_down.weight
     24:   58720256 |  4096, 14336,     1,     1 | F16     | blk.10.ffn_gate.weight
     25:   58720256 |  4096, 14336,     1,     1 | F16     | blk.10.ffn_up.weight
     26:       4096 |  4096,     1,     1,     1 | F32     | blk.10.ffn_norm.weight
     27:    4194304 |  4096,  1024,     1,     1 | F16     | blk.10.attn_k.weight
     28:   16777216 |  4096,  4096,     1,     1 | F16     | blk.10.attn_output.weight
     29:   16777216 |  4096,  4096,     1,     1 | F16     | blk.10.attn_q.weight
     30:         64 |    64,     1,     1,     1 | F32     | blk.10.attn_rot_embd
     31:    4194304 |  4096,  1024,     1,     1 | F16     | blk.10.attn_v.weight
     32:       4096 |  4096,     1,     1,     1 | F32     | blk.11.attn_norm.weight
     33:   58720256 | 14336,  4096,     1,     1 | F16     | blk.11.ffn_down.weight
     34:   58720256 |  4096, 14336,     1,     1 | F16     | blk.11.ffn_gate.weight
     35:   58720256 |  4096, 14336,     1,     1 | F16     | blk.11.ffn_up.weight
     36:       4096 |  4096,     1,     1,     1 | F32     | blk.11.ffn_norm.weight
     37:    4194304 |  4096,  1024,     1,     1 | F16     | blk.11.attn_k.weight
     38:   16777216 |  4096,  4096,     1,     1 | F16     | blk.11.attn_output.weight
     39:   16777216 |  4096,  4096,     1,     1 | F16     | blk.11.attn_q.weight
     40:         64 |    64,     1,     1,     1 | F32     | blk.11.attn_rot_embd
     41:    4194304 |  4096,  1024,     1,     1 | F16     | blk.11.attn_v.weight
     42:       4096 |  4096,     1,     1,     1 | F32     | blk.12.attn_norm.weight
     43:   58720256 | 14336,  4096,     1,     1 | F16     | blk.12.ffn_down.weight
     44:   58720256 |  4096, 14336,     1,     1 | F16     | blk.12.ffn_gate.weight
     45:   58720256 |  4096, 14336,     1,     1 | F16     | blk.12.ffn_up.weight
     46:       4096 |  4096,     1,     1,     1 | F32     | blk.12.ffn_norm.weight
     47:    4194304 |  4096,  1024,     1,     1 | F16     | blk.12.attn_k.weight
     48:   16777216 |  4096,  4096,     1,     1 | F16     | blk.12.attn_output.weight
     49:   16777216 |  4096,  4096,     1,     1 | F16     | blk.12.attn_q.weight
     50:         64 |    64,     1,     1,     1 | F32     | blk.12.attn_rot_embd
     51:    4194304 |  4096,  1024,     1,     1 | F16     | blk.12.attn_v.weight
     52:       4096 |  4096,     1,     1,     1 | F32     | blk.13.attn_norm.weight
     53:   58720256 | 14336,  4096,     1,     1 | F16     | blk.13.ffn_down.weight
     54:   58720256 |  4096, 14336,     1,     1 | F16     | blk.13.ffn_gate.weight
     55:   58720256 |  4096, 14336,     1,     1 | F16     | blk.13.ffn_up.weight
     56:       4096 |  4096,     1,     1,     1 | F32     | blk.13.ffn_norm.weight
     57:    4194304 |  4096,  1024,     1,     1 | F16     | blk.13.attn_k.weight
     58:   16777216 |  4096,  4096,     1,     1 | F16     | blk.13.attn_output.weight
     59:   16777216 |  4096,  4096,     1,     1 | F16     | blk.13.attn_q.weight
     60:         64 |    64,     1,     1,     1 | F32     | blk.13.attn_rot_embd
     61:    4194304 |  4096,  1024,     1,     1 | F16     | blk.13.attn_v.weight
     62:       4096 |  4096,     1,     1,     1 | F32     | blk.14.attn_norm.weight
     63:   58720256 | 14336,  4096,     1,     1 | F16     | blk.14.ffn_down.weight
     64:   58720256 |  4096, 14336,     1,     1 | F16     | blk.14.ffn_gate.weight
     65:   58720256 |  4096, 14336,     1,     1 | F16     | blk.14.ffn_up.weight
     66:       4096 |  4096,     1,     1,     1 | F32     | blk.14.ffn_norm.weight
     67:    4194304 |  4096,  1024,     1,     1 | F16     | blk.14.attn_k.weight
     68:   16777216 |  4096,  4096,     1,     1 | F16     | blk.14.attn_output.weight
     69:   16777216 |  4096,  4096,     1,     1 | F16     | blk.14.attn_q.weight
     70:         64 |    64,     1,     1,     1 | F32     | blk.14.attn_rot_embd
     71:    4194304 |  4096,  1024,     1,     1 | F16     | blk.14.attn_v.weight
     72:       4096 |  4096,     1,     1,     1 | F32     | blk.15.attn_norm.weight
     73:   58720256 | 14336,  4096,     1,     1 | F16     | blk.15.ffn_down.weight
     74:   58720256 |  4096, 14336,     1,     1 | F16     | blk.15.ffn_gate.weight
     75:   58720256 |  4096, 14336,     1,     1 | F16     | blk.15.ffn_up.weight
     76:       4096 |  4096,     1,     1,     1 | F32     | blk.15.ffn_norm.weight
     77:    4194304 |  4096,  1024,     1,     1 | F16     | blk.15.attn_k.weight
     78:   16777216 |  4096,  4096,     1,     1 | F16     | blk.15.attn_output.weight
     79:   16777216 |  4096,  4096,     1,     1 | F16     | blk.15.attn_q.weight
     80:         64 |    64,     1,     1,     1 | F32     | blk.15.attn_rot_embd
     81:    4194304 |  4096,  1024,     1,     1 | F16     | blk.15.attn_v.weight
     82:       4096 |  4096,     1,     1,     1 | F32     | blk.16.attn_norm.weight
     83:   58720256 | 14336,  4096,     1,     1 | F16     | blk.16.ffn_down.weight
     84:   58720256 |  4096, 14336,     1,     1 | F16     | blk.16.ffn_gate.weight
     85:   58720256 |  4096, 14336,     1,     1 | F16     | blk.16.ffn_up.weight
     86:       4096 |  4096,     1,     1,     1 | F32     | blk.16.ffn_norm.weight
     87:    4194304 |  4096,  1024,     1,     1 | F16     | blk.16.attn_k.weight
     88:   16777216 |  4096,  4096,     1,     1 | F16     | blk.16.attn_output.weight
     89:   16777216 |  4096,  4096,     1,     1 | F16     | blk.16.attn_q.weight
     90:         64 |    64,     1,     1,     1 | F32     | blk.16.attn_rot_embd
     91:    4194304 |  4096,  1024,     1,     1 | F16     | blk.16.attn_v.weight
     92:       4096 |  4096,     1,     1,     1 | F32     | blk.17.attn_norm.weight
     93:   58720256 | 14336,  4096,     1,     1 | F16     | blk.17.ffn_down.weight
     94:   58720256 |  4096, 14336,     1,     1 | F16     | blk.17.ffn_gate.weight
     95:   58720256 |  4096, 14336,     1,     1 | F16     | blk.17.ffn_up.weight
     96:       4096 |  4096,     1,     1,     1 | F32     | blk.17.ffn_norm.weight
     97:    4194304 |  4096,  1024,     1,     1 | F16     | blk.17.attn_k.weight
     98:   16777216 |  4096,  4096,     1,     1 | F16     | blk.17.attn_output.weight
     99:   16777216 |  4096,  4096,     1,     1 | F16     | blk.17.attn_q.weight
    100:         64 |    64,     1,     1,     1 | F32     | blk.17.attn_rot_embd
    101:    4194304 |  4096,  1024,     1,     1 | F16     | blk.17.attn_v.weight
    102:       4096 |  4096,     1,     1,     1 | F32     | blk.18.attn_norm.weight
    103:   58720256 | 14336,  4096,     1,     1 | F16     | blk.18.ffn_down.weight
    104:   58720256 |  4096, 14336,     1,     1 | F16     | blk.18.ffn_gate.weight
    105:   58720256 |  4096, 14336,     1,     1 | F16     | blk.18.ffn_up.weight
    106:       4096 |  4096,     1,     1,     1 | F32     | blk.18.ffn_norm.weight
    107:    4194304 |  4096,  1024,     1,     1 | F16     | blk.18.attn_k.weight
    108:   16777216 |  4096,  4096,     1,     1 | F16     | blk.18.attn_output.weight
    109:   16777216 |  4096,  4096,     1,     1 | F16     | blk.18.attn_q.weight
    110:         64 |    64,     1,     1,     1 | F32     | blk.18.attn_rot_embd
    111:    4194304 |  4096,  1024,     1,     1 | F16     | blk.18.attn_v.weight
    112:       4096 |  4096,     1,     1,     1 | F32     | blk.19.attn_norm.weight
    113:   58720256 | 14336,  4096,     1,     1 | F16     | blk.19.ffn_down.weight
    114:   58720256 |  4096, 14336,     1,     1 | F16     | blk.19.ffn_gate.weight
    115:   58720256 |  4096, 14336,     1,     1 | F16     | blk.19.ffn_up.weight
    116:       4096 |  4096,     1,     1,     1 | F32     | blk.19.ffn_norm.weight
    117:    4194304 |  4096,  1024,     1,     1 | F16     | blk.19.attn_k.weight
    118:   16777216 |  4096,  4096,     1,     1 | F16     | blk.19.attn_output.weight
    119:   16777216 |  4096,  4096,     1,     1 | F16     | blk.19.attn_q.weight
    120:         64 |    64,     1,     1,     1 | F32     | blk.19.attn_rot_embd
    121:    4194304 |  4096,  1024,     1,     1 | F16     | blk.19.attn_v.weight
    122:       4096 |  4096,     1,     1,     1 | F32     | blk.2.attn_norm.weight
    123:   58720256 | 14336,  4096,     1,     1 | F16     | blk.2.ffn_down.weight
    124:   58720256 |  4096, 14336,     1,     1 | F16     | blk.2.ffn_gate.weight
    125:   58720256 |  4096, 14336,     1,     1 | F16     | blk.2.ffn_up.weight
    126:       4096 |  4096,     1,     1,     1 | F32     | blk.2.ffn_norm.weight
    127:    4194304 |  4096,  1024,     1,     1 | F16     | blk.2.attn_k.weight
    128:   16777216 |  4096,  4096,     1,     1 | F16     | blk.2.attn_output.weight
    129:   16777216 |  4096,  4096,     1,     1 | F16     | blk.2.attn_q.weight
    130:         64 |    64,     1,     1,     1 | F32     | blk.2.attn_rot_embd
    131:    4194304 |  4096,  1024,     1,     1 | F16     | blk.2.attn_v.weight
    132:       4096 |  4096,     1,     1,     1 | F32     | blk.20.attn_norm.weight
    133:   58720256 |  4096, 14336,     1,     1 | F16     | blk.20.ffn_gate.weight
    134:       4096 |  4096,     1,     1,     1 | F32     | blk.20.ffn_norm.weight
    135:    4194304 |  4096,  1024,     1,     1 | F16     | blk.20.attn_k.weight
    136:   16777216 |  4096,  4096,     1,     1 | F16     | blk.20.attn_output.weight
    137:   16777216 |  4096,  4096,     1,     1 | F16     | blk.20.attn_q.weight
    138:         64 |    64,     1,     1,     1 | F32     | blk.20.attn_rot_embd
    139:    4194304 |  4096,  1024,     1,     1 | F16     | blk.20.attn_v.weight
    140:       4096 |  4096,     1,     1,     1 | F32     | blk.3.attn_norm.weight
    141:   58720256 | 14336,  4096,     1,     1 | F16     | blk.3.ffn_down.weight
    142:   58720256 |  4096, 14336,     1,     1 | F16     | blk.3.ffn_gate.weight
    143:   58720256 |  4096, 14336,     1,     1 | F16     | blk.3.ffn_up.weight
    144:       4096 |  4096,     1,     1,     1 | F32     | blk.3.ffn_norm.weight
    145:    4194304 |  4096,  1024,     1,     1 | F16     | blk.3.attn_k.weight
    146:   16777216 |  4096,  4096,     1,     1 | F16     | blk.3.attn_output.weight
    147:   16777216 |  4096,  4096,     1,     1 | F16     | blk.3.attn_q.weight
    148:         64 |    64,     1,     1,     1 | F32     | blk.3.attn_rot_embd
    149:    4194304 |  4096,  1024,     1,     1 | F16     | blk.3.attn_v.weight
    150:       4096 |  4096,     1,     1,     1 | F32     | blk.4.attn_norm.weight
    151:   58720256 | 14336,  4096,     1,     1 | F16     | blk.4.ffn_down.weight
    152:   58720256 |  4096, 14336,     1,     1 | F16     | blk.4.ffn_gate.weight
    153:   58720256 |  4096, 14336,     1,     1 | F16     | blk.4.ffn_up.weight
    154:       4096 |  4096,     1,     1,     1 | F32     | blk.4.ffn_norm.weight
    155:    4194304 |  4096,  1024,     1,     1 | F16     | blk.4.attn_k.weight
    156:   16777216 |  4096,  4096,     1,     1 | F16     | blk.4.attn_output.weight
    157:   16777216 |  4096,  4096,     1,     1 | F16     | blk.4.attn_q.weight
    158:         64 |    64,     1,     1,     1 | F32     | blk.4.attn_rot_embd
    159:    4194304 |  4096,  1024,     1,     1 | F16     | blk.4.attn_v.weight
    160:       4096 |  4096,     1,     1,     1 | F32     | blk.5.attn_norm.weight
    161:   58720256 | 14336,  4096,     1,     1 | F16     | blk.5.ffn_down.weight
    162:   58720256 |  4096, 14336,     1,     1 | F16     | blk.5.ffn_gate.weight
    163:   58720256 |  4096, 14336,     1,     1 | F16     | blk.5.ffn_up.weight
    164:       4096 |  4096,     1,     1,     1 | F32     | blk.5.ffn_norm.weight
    165:    4194304 |  4096,  1024,     1,     1 | F16     | blk.5.attn_k.weight
    166:   16777216 |  4096,  4096,     1,     1 | F16     | blk.5.attn_output.weight
    167:   16777216 |  4096,  4096,     1,     1 | F16     | blk.5.attn_q.weight
    168:         64 |    64,     1,     1,     1 | F32     | blk.5.attn_rot_embd
    169:    4194304 |  4096,  1024,     1,     1 | F16     | blk.5.attn_v.weight
    170:       4096 |  4096,     1,     1,     1 | F32     | blk.6.attn_norm.weight
    171:   58720256 | 14336,  4096,     1,     1 | F16     | blk.6.ffn_down.weight
    172:   58720256 |  4096, 14336,     1,     1 | F16     | blk.6.ffn_gate.weight
    173:   58720256 |  4096, 14336,     1,     1 | F16     | blk.6.ffn_up.weight
    174:       4096 |  4096,     1,     1,     1 | F32     | blk.6.ffn_norm.weight
    175:    4194304 |  4096,  1024,     1,     1 | F16     | blk.6.attn_k.weight
    176:   16777216 |  4096,  4096,     1,     1 | F16     | blk.6.attn_output.weight
    177:   16777216 |  4096,  4096,     1,     1 | F16     | blk.6.attn_q.weight
    178:         64 |    64,     1,     1,     1 | F32     | blk.6.attn_rot_embd
    179:    4194304 |  4096,  1024,     1,     1 | F16     | blk.6.attn_v.weight
    180:       4096 |  4096,     1,     1,     1 | F32     | blk.7.attn_norm.weight
    181:   58720256 | 14336,  4096,     1,     1 | F16     | blk.7.ffn_down.weight
    182:   58720256 |  4096, 14336,     1,     1 | F16     | blk.7.ffn_gate.weight
    183:   58720256 |  4096, 14336,     1,     1 | F16     | blk.7.ffn_up.weight
    184:       4096 |  4096,     1,     1,     1 | F32     | blk.7.ffn_norm.weight
    185:    4194304 |  4096,  1024,     1,     1 | F16     | blk.7.attn_k.weight
    186:   16777216 |  4096,  4096,     1,     1 | F16     | blk.7.attn_output.weight
    187:   16777216 |  4096,  4096,     1,     1 | F16     | blk.7.attn_q.weight
    188:         64 |    64,     1,     1,     1 | F32     | blk.7.attn_rot_embd
    189:    4194304 |  4096,  1024,     1,     1 | F16     | blk.7.attn_v.weight
    190:       4096 |  4096,     1,     1,     1 | F32     | blk.8.attn_norm.weight
    191:   58720256 | 14336,  4096,     1,     1 | F16     | blk.8.ffn_down.weight
    192:   58720256 |  4096, 14336,     1,     1 | F16     | blk.8.ffn_gate.weight
    193:   58720256 |  4096, 14336,     1,     1 | F16     | blk.8.ffn_up.weight
    194:       4096 |  4096,     1,     1,     1 | F32     | blk.8.ffn_norm.weight
    195:    4194304 |  4096,  1024,     1,     1 | F16     | blk.8.attn_k.weight
    196:   16777216 |  4096,  4096,     1,     1 | F16     | blk.8.attn_output.weight
    197:   16777216 |  4096,  4096,     1,     1 | F16     | blk.8.attn_q.weight
    198:         64 |    64,     1,     1,     1 | F32     | blk.8.attn_rot_embd
    199:    4194304 |  4096,  1024,     1,     1 | F16     | blk.8.attn_v.weight
    200:       4096 |  4096,     1,     1,     1 | F32     | blk.9.attn_norm.weight
    201:   58720256 | 14336,  4096,     1,     1 | F16     | blk.9.ffn_down.weight
    202:   58720256 |  4096, 14336,     1,     1 | F16     | blk.9.ffn_gate.weight
    203:   58720256 |  4096, 14336,     1,     1 | F16     | blk.9.ffn_up.weight
    204:       4096 |  4096,     1,     1,     1 | F32     | blk.9.ffn_norm.weight
    205:    4194304 |  4096,  1024,     1,     1 | F16     | blk.9.attn_k.weight
    206:   16777216 |  4096,  4096,     1,     1 | F16     | blk.9.attn_output.weight
    207:   16777216 |  4096,  4096,     1,     1 | F16     | blk.9.attn_q.weight
    208:         64 |    64,     1,     1,     1 | F32     | blk.9.attn_rot_embd
    209:    4194304 |  4096,  1024,     1,     1 | F16     | blk.9.attn_v.weight
    210:  525336576 |  4096, 128256,     1,     1 | F16     | output.weight
    211:   58720256 | 14336,  4096,     1,     1 | F16     | blk.20.ffn_down.weight
    212:   58720256 |  4096, 14336,     1,     1 | F16     | blk.20.ffn_up.weight
    213:       4096 |  4096,     1,     1,     1 | F32     | blk.21.attn_norm.weight
    214:   58720256 | 14336,  4096,     1,     1 | F16     | blk.21.ffn_down.weight
    215:   58720256 |  4096, 14336,     1,     1 | F16     | blk.21.ffn_gate.weight
    216:   58720256 |  4096, 14336,     1,     1 | F16     | blk.21.ffn_up.weight
    217:       4096 |  4096,     1,     1,     1 | F32     | blk.21.ffn_norm.weight
    218:    4194304 |  4096,  1024,     1,     1 | F16     | blk.21.attn_k.weight
    219:   16777216 |  4096,  4096,     1,     1 | F16     | blk.21.attn_output.weight
    220:   16777216 |  4096,  4096,     1,     1 | F16     | blk.21.attn_q.weight
    221:         64 |    64,     1,     1,     1 | F32     | blk.21.attn_rot_embd
    222:    4194304 |  4096,  1024,     1,     1 | F16     | blk.21.attn_v.weight
    223:       4096 |  4096,     1,     1,     1 | F32     | blk.22.attn_norm.weight
    224:   58720256 | 14336,  4096,     1,     1 | F16     | blk.22.ffn_down.weight
    225:   58720256 |  4096, 14336,     1,     1 | F16     | blk.22.ffn_gate.weight
    226:   58720256 |  4096, 14336,     1,     1 | F16     | blk.22.ffn_up.weight
    227:       4096 |  4096,     1,     1,     1 | F32     | blk.22.ffn_norm.weight
    228:    4194304 |  4096,  1024,     1,     1 | F16     | blk.22.attn_k.weight
    229:   16777216 |  4096,  4096,     1,     1 | F16     | blk.22.attn_output.weight
    230:   16777216 |  4096,  4096,     1,     1 | F16     | blk.22.attn_q.weight
    231:         64 |    64,     1,     1,     1 | F32     | blk.22.attn_rot_embd
    232:    4194304 |  4096,  1024,     1,     1 | F16     | blk.22.attn_v.weight
    233:       4096 |  4096,     1,     1,     1 | F32     | blk.23.attn_norm.weight
    234:   58720256 | 14336,  4096,     1,     1 | F16     | blk.23.ffn_down.weight
    235:   58720256 |  4096, 14336,     1,     1 | F16     | blk.23.ffn_gate.weight
    236:   58720256 |  4096, 14336,     1,     1 | F16     | blk.23.ffn_up.weight
    237:       4096 |  4096,     1,     1,     1 | F32     | blk.23.ffn_norm.weight
    238:    4194304 |  4096,  1024,     1,     1 | F16     | blk.23.attn_k.weight
    239:   16777216 |  4096,  4096,     1,     1 | F16     | blk.23.attn_output.weight
    240:   16777216 |  4096,  4096,     1,     1 | F16     | blk.23.attn_q.weight
    241:         64 |    64,     1,     1,     1 | F32     | blk.23.attn_rot_embd
    242:    4194304 |  4096,  1024,     1,     1 | F16     | blk.23.attn_v.weight
    243:       4096 |  4096,     1,     1,     1 | F32     | blk.24.attn_norm.weight
    244:   58720256 | 14336,  4096,     1,     1 | F16     | blk.24.ffn_down.weight
    245:   58720256 |  4096, 14336,     1,     1 | F16     | blk.24.ffn_gate.weight
    246:   58720256 |  4096, 14336,     1,     1 | F16     | blk.24.ffn_up.weight
    247:       4096 |  4096,     1,     1,     1 | F32     | blk.24.ffn_norm.weight
    248:    4194304 |  4096,  1024,     1,     1 | F16     | blk.24.attn_k.weight
    249:   16777216 |  4096,  4096,     1,     1 | F16     | blk.24.attn_output.weight
    250:   16777216 |  4096,  4096,     1,     1 | F16     | blk.24.attn_q.weight
    251:         64 |    64,     1,     1,     1 | F32     | blk.24.attn_rot_embd
    252:    4194304 |  4096,  1024,     1,     1 | F16     | blk.24.attn_v.weight
    253:       4096 |  4096,     1,     1,     1 | F32     | blk.25.attn_norm.weight
    254:   58720256 | 14336,  4096,     1,     1 | F16     | blk.25.ffn_down.weight
    255:   58720256 |  4096, 14336,     1,     1 | F16     | blk.25.ffn_gate.weight
    256:   58720256 |  4096, 14336,     1,     1 | F16     | blk.25.ffn_up.weight
    257:       4096 |  4096,     1,     1,     1 | F32     | blk.25.ffn_norm.weight
    258:    4194304 |  4096,  1024,     1,     1 | F16     | blk.25.attn_k.weight
    259:   16777216 |  4096,  4096,     1,     1 | F16     | blk.25.attn_output.weight
    260:   16777216 |  4096,  4096,     1,     1 | F16     | blk.25.attn_q.weight
    261:         64 |    64,     1,     1,     1 | F32     | blk.25.attn_rot_embd
    262:    4194304 |  4096,  1024,     1,     1 | F16     | blk.25.attn_v.weight
    263:       4096 |  4096,     1,     1,     1 | F32     | blk.26.attn_norm.weight
    264:   58720256 | 14336,  4096,     1,     1 | F16     | blk.26.ffn_down.weight
    265:   58720256 |  4096, 14336,     1,     1 | F16     | blk.26.ffn_gate.weight
    266:   58720256 |  4096, 14336,     1,     1 | F16     | blk.26.ffn_up.weight
    267:       4096 |  4096,     1,     1,     1 | F32     | blk.26.ffn_norm.weight
    268:    4194304 |  4096,  1024,     1,     1 | F16     | blk.26.attn_k.weight
    269:   16777216 |  4096,  4096,     1,     1 | F16     | blk.26.attn_output.weight
    270:   16777216 |  4096,  4096,     1,     1 | F16     | blk.26.attn_q.weight
    271:         64 |    64,     1,     1,     1 | F32     | blk.26.attn_rot_embd
    272:    4194304 |  4096,  1024,     1,     1 | F16     | blk.26.attn_v.weight
    273:       4096 |  4096,     1,     1,     1 | F32     | blk.27.attn_norm.weight
    274:   58720256 | 14336,  4096,     1,     1 | F16     | blk.27.ffn_down.weight
    275:   58720256 |  4096, 14336,     1,     1 | F16     | blk.27.ffn_gate.weight
    276:   58720256 |  4096, 14336,     1,     1 | F16     | blk.27.ffn_up.weight
    277:       4096 |  4096,     1,     1,     1 | F32     | blk.27.ffn_norm.weight
    278:    4194304 |  4096,  1024,     1,     1 | F16     | blk.27.attn_k.weight
    279:   16777216 |  4096,  4096,     1,     1 | F16     | blk.27.attn_output.weight
    280:   16777216 |  4096,  4096,     1,     1 | F16     | blk.27.attn_q.weight
    281:         64 |    64,     1,     1,     1 | F32     | blk.27.attn_rot_embd
    282:    4194304 |  4096,  1024,     1,     1 | F16     | blk.27.attn_v.weight
    283:       4096 |  4096,     1,     1,     1 | F32     | blk.28.attn_norm.weight
    284:   58720256 | 14336,  4096,     1,     1 | F16     | blk.28.ffn_down.weight
    285:   58720256 |  4096, 14336,     1,     1 | F16     | blk.28.ffn_gate.weight
    286:   58720256 |  4096, 14336,     1,     1 | F16     | blk.28.ffn_up.weight
    287:       4096 |  4096,     1,     1,     1 | F32     | blk.28.ffn_norm.weight
    288:    4194304 |  4096,  1024,     1,     1 | F16     | blk.28.attn_k.weight
    289:   16777216 |  4096,  4096,     1,     1 | F16     | blk.28.attn_output.weight
    290:   16777216 |  4096,  4096,     1,     1 | F16     | blk.28.attn_q.weight
    291:         64 |    64,     1,     1,     1 | F32     | blk.28.attn_rot_embd
    292:    4194304 |  4096,  1024,     1,     1 | F16     | blk.28.attn_v.weight
    293:       4096 |  4096,     1,     1,     1 | F32     | blk.29.attn_norm.weight
    294:   58720256 | 14336,  4096,     1,     1 | F16     | blk.29.ffn_down.weight
    295:   58720256 |  4096, 14336,     1,     1 | F16     | blk.29.ffn_gate.weight
    296:   58720256 |  4096, 14336,     1,     1 | F16     | blk.29.ffn_up.weight
    297:       4096 |  4096,     1,     1,     1 | F32     | blk.29.ffn_norm.weight
    298:    4194304 |  4096,  1024,     1,     1 | F16     | blk.29.attn_k.weight
    299:   16777216 |  4096,  4096,     1,     1 | F16     | blk.29.attn_output.weight
    300:   16777216 |  4096,  4096,     1,     1 | F16     | blk.29.attn_q.weight
    301:         64 |    64,     1,     1,     1 | F32     | blk.29.attn_rot_embd
    302:    4194304 |  4096,  1024,     1,     1 | F16     | blk.29.attn_v.weight
    303:       4096 |  4096,     1,     1,     1 | F32     | blk.30.attn_norm.weight
    304:   58720256 | 14336,  4096,     1,     1 | F16     | blk.30.ffn_down.weight
    305:   58720256 |  4096, 14336,     1,     1 | F16     | blk.30.ffn_gate.weight
    306:   58720256 |  4096, 14336,     1,     1 | F16     | blk.30.ffn_up.weight
    307:       4096 |  4096,     1,     1,     1 | F32     | blk.30.ffn_norm.weight
    308:    4194304 |  4096,  1024,     1,     1 | F16     | blk.30.attn_k.weight
    309:   16777216 |  4096,  4096,     1,     1 | F16     | blk.30.attn_output.weight
    310:   16777216 |  4096,  4096,     1,     1 | F16     | blk.30.attn_q.weight
    311:         64 |    64,     1,     1,     1 | F32     | blk.30.attn_rot_embd
    312:    4194304 |  4096,  1024,     1,     1 | F16     | blk.30.attn_v.weight
    313:       4096 |  4096,     1,     1,     1 | F32     | blk.31.attn_norm.weight
    314:   58720256 | 14336,  4096,     1,     1 | F16     | blk.31.ffn_down.weight
    315:   58720256 |  4096, 14336,     1,     1 | F16     | blk.31.ffn_gate.weight
    316:   58720256 |  4096, 14336,     1,     1 | F16     | blk.31.ffn_up.weight
    317:       4096 |  4096,     1,     1,     1 | F32     | blk.31.ffn_norm.weight
    318:    4194304 |  4096,  1024,     1,     1 | F16     | blk.31.attn_k.weight
    319:   16777216 |  4096,  4096,     1,     1 | F16     | blk.31.attn_output.weight
    320:   16777216 |  4096,  4096,     1,     1 | F16     | blk.31.attn_q.weight
    321:         64 |    64,     1,     1,     1 | F32     | blk.31.attn_rot_embd
    322:    4194304 |  4096,  1024,     1,     1 | F16     | blk.31.attn_v.weight
    323:       4096 |  4096,     1,     1,     1 | F32     | output_norm.weight

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions